Ghazaryan, Gohar: Analysis of Land Surface Dynamics in Ukraine Observed by Satellite Sensors. - Bonn, 2020. - Dissertation, Rheinische Friedrich-Wilhelms-Universität Bonn.
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5-58674
@phdthesis{handle:20.500.11811/8413,
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5-58674,
author = {{Gohar Ghazaryan}},
title = {Analysis of Land Surface Dynamics in Ukraine Observed by Satellite Sensors},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2020,
month = jun,

note = {Land surface changes, induced by anthropogenic or climatic drivers, can dramatically impact ecosystem functioning. The growing amount of data from remote sensing and complementary data sources greatly supports the quantification of land surface changes. This research aimed to use several remotely sensed datasets to explore inter-annual and seasonal variability of land surface in Ukraine at multiple spatial scales. The country was chosen as a study area, as it has experienced immense institutional and environmental changes during recent decades.
For the analysis at country scale, the main aim was to understand land surface dynamics and to assess processes underlying the changes. For this, Global Inventory Modeling and Mapping Studies (GIMMS) Normalized Difference Vegetation Index (NDVI) time series were used. Abrupt and gradual changes were delineated, and the relationships of land surface changes and climatic variables were addressed. Among the factors analyzed, air temperature explained the largest portion of NDVI variability.
For the analysis at local scale the focus was put on crop identification and crop condition monitoring. For crop mapping, the combined use of time series observations derived from Landsat-8 and Sentinel-1 was tested. The phenology was modeled by fitting harmonic function and training samples were generated based on the fit. Three classification algorithms (support vector machines, random forest, decision fusion) were tested for crop mapping. Overall classification accuracies exceeded 80% for random forest and decision fusion when using Landsat and Sentinel based seasonal composites.
For drought impact monitoring in croplands, time series from optical (Landsat, MODIS, Sentinel- 2) and Synthetic Aperture Radar (SAR) data was used. Indicators were derived based on optical and Sentinel-1 data. Logistic regression was used to evaluate the drought-induced variability of remotely sensed parameters. Growing season maximum NDMI and NDVI and SAR backscatter reflect the impact of agricultural drought. Land Surface Temperature (LST) was also a useful indicator of crop condition, especially for maize and sunflower, with prediction rates of 86% and 71%, respectively.
Furthermore, to contribute to not only remotely sensed data analysis but also their dissemination, a web application was developed that enables the provision of customizable geospatial tools and products. The user is able to define either spatial or temporal parameters (or both), change the used algorithms (e.g. change detection, anomaly detection) or visualization parameters based on the preferred data sources and get access to previously discussed outputs.
The research for this thesis combined different trend analysis techniques, integrated multiple datasets, and advanced statistical modeling at different scales. This allowed analyses to go beyond descriptive information like overall vegetation status and dynamics, land degradation or crop stress; but to derive valuable spatially explicit information towards a better understanding of change drivers. This information forms the essential basis for advanced models and leads the way to better decision making for sustainable land management.},

url = {https://hdl.handle.net/20.500.11811/8413}
}

Die folgenden Nutzungsbestimmungen sind mit dieser Ressource verbunden:

InCopyright